Using crowdsourcing to compare temporal, social temporal, and probability discounting among obese and non-obese individuals

Highlights

• We recruited a large sample (n = 1181) of participants using crowdsourcing technology.

• We assessed discount rates for delayed, probabilistic, and social monetary rewards.

• Delay and social discounting rates were higher in obese than non-obese individuals.

• Probability discounting rates were comparable in obese than non-obese individuals.

Abstract

Previous research comparing obese and non-obese samples on the delayed discounting procedure has produced mixed results. The aim of the current study was to clarify these discrepant findings by comparing a variety of temporal discounting measures in a large sample of internet users (n = 1163) obtained from a crowdsourcing service, Amazon Mechanical Turk (AMT). Measures of temporal, social–temporal (a combination of standard and social temporal), and probability discounting were obtained. Significant differences were obtained on all discounting measures except probability discounting, but the obtained effect sizes were small. These data suggest that larger-N studies will be more likely to detect differences between obese and non-obese samples, and may afford the opportunity, in future studies, to decompose a large obese sample into different subgroups to examine the effect of other relevant measures, such as the reinforcing value of food, on discounting.

Introduction

Obesity is a major public health problem that impacts approximately 36% of American adults (CDC, 2013). Several conceptualizations of obesity suggest that the failure of self-control, sometimes called impulsivity or executive dysfunction, plays a major role in the genesis and maintenance of weight control problems (Epstein, Salvy, Carr, Dearing, & Bickel, 2010). Specifically, the immediate impulse to consume food prevails over the future health goals, which can be obtained by engaging in fewer and smaller bouts of eating. From behavioral and neuroeconomic perspectives, this failure of self-control results from normal learning mechanisms bring commandeered by the pathological processing of food rewards, further this dysfunction becomes exacerbated over time (Bickel et al., 2007). These pathological valuations may distort decision making by altering the regulatory balance between two neurobehavioral decision systems, resulting in an over-valuation of immediate commodities (e.g., food), and an under valuation of longer-term commodities (e.g., improved health) (Bickel et al., 2012, Bickel et al., 2011, Carr et al., 2011). Importantly, delay discounting, which measures the decline in the value of a reinforcer as a function of the delay to its receipt, reflects the regulatory balance between impulsive (i.e., limbic and paralimbic areas) and executive (prefrontal cortex) neurobehavioral systems (Bickel et al., 2007). Individuals suffering from a variety of behavioral disorders (e.g., drug addiction, pathological gambling, etc.) excessively discount delayed rewards, suggesting that excessive discounting may function as a trans-disease process (Bickel, Jarmolowicz, Mueller, Koffarnus, & Gatchalian, 2012).

Delay discounting has been shown to be an index of food reinforcement (Epstein et al., 2010), and this is true on both the physiological and behavioral level, however the few studies that have used delay discounting to research the eating behavior of obese subjects have produced inconclusive results. Physiologically, discounting in those who engage in excessive hedonic eating is thought to be a manifestation of systemic dopaminergic dysfunction (see Appelhans, 2009, for a review). Further, activity in regions of the prefrontal cortex that were more active on difficult delayed discounting trials were predictive of weight gain (within the next two years) in obese subjects (Kishinevsky et al., 2012), however a follow-up study on this same group of subjects failed find a correlation between current BMI measures and discounting (Stoeckel, Murdaugh, Cox, Cook, & Weller, 2013). Behaviorally, discounting has been shown to be an index of excessive eating and resulting poor health across a number of different dimensions. Obese women, but not men, were shown to discount more than normal weight women, normal weight men, and obese men (Weller, Cook, Avsar, & Cox, 2008). Likewise, obese adolescent smokers have been found to discount more than normal-weight smokers (Fields, Sabet, Peal, & Reynolds, 2011). Similarly, normal-weight adolescents discount less than overweight and obese adolescents, and overweight adolescents discount less than obese adolescents (Fields, Sabet, & Reynolds, 2013). Obese and overweight women with greater discounting order more high-energy foods, and consumed more calories, when eating away-from home meals, but this is not true of meals prepared at home (Appelhans et al., 2012). Finally, both male and female subjects with higher percent body fat have greater discounting for food rewards (Hendrickson and Rasmussen, 2013, Rasmussen et al., 2010), but only female participants with higher percent body fat show greater discounting for monetary rewards (Rasmussen et al., 2010).

The study of discounting in the obese so far has not examined a relatively new variant of the delayed discounting procedure referred to as social temporal discounting (Bickel et al., 2012, Bickel et al., 2012, Bickel et al., 2012, Charlton et al., 2013). Social temporal discounting provides a measure of the decline in value of a reinforcer that will be distributed to a group of individuals (including the participant) as a function of the delay to the reinforcer’s receipt (Bickel, Jarmolowicz, Mueller, Franck, et al., 2012). This procedure presents participants with choices between a reinforcer that will be evenly distributed among a group of unspecified individuals (including the participant) now vs. a larger reward to be shared among that same group later (i.e., “we now” vs. “we later”; Charlton et al., 2013). Comparison of standard delay discounting (i.e., “me now” vs. “me later”) and social temporal discounting among college students resulted in greater preference for the delayed option (less discounting) in the social temporal option relative to preference in the standard discounting procedure. Moreover, a study examining temporal discounting, social temporal discounting, and a combination of standard temporal discounting and social temporal discounting (i.e., “me now” vs. “we later”) in problem drinkers showed comparable results to controls on the extent of discounting on social temporal and the combination discounting task, while smokers discounted significantly more on all forms of discounting relative to controls (Bickel et al., 2012, Bickel et al., 2012, Bickel et al., 2012, Bickel et al., 2012). Thus, whether social temporal discounting engenders more or less preference for the delayed option is, in part, dependent on the population under study.

In this study, we obtained a large sample of participants (n = 1181) by using Amazon’s Mechanical Turk (AMT). AMT, a crowdsourcing service, permits researchers to post tasks or questions which are then answered by a potential participant pool of more than 500,000 potential research volunteers to complete (The Economist, 2012). This participant pool is considerably more diverse than typical university-based samples (Buhrmester et al., 2011, Jarmolowicz et al., 2012, The Economist, 2012), and studies using AMT as a source of participants have replicated previously reported findings (Bickel et al., 2012, Sprouse, 2011). The present study used AMT to collect data from a large sample to clarify the extent of differences in temporal, social temporal discounting, combination of standard and social temporal discounting, and probability discounting between obese and control participants.